Phase 5: 3D Generation · 45 min · Python · Tripo API

3D Generation Modes

Text-to-3D for concepts, image-to-3D for products, multi-view for photorealism — each mode serves a different production need.

Hiring signal: Knowing all 3D generation modes and when to use each demonstrates practical 3D generation engineering skills.

What you will learn

Implement text-to-3D: generate 3D model from text description
Implement image-to-3D: generate 3D model from a single or multiple 2D images
Understand multi-view reconstruction: multiple angles → 3D model
Understand Gaussian Splatting: radiance field representation for photorealistic 3D

The Problem

A team needs a 3D model of a specific product (a red coffee mug with a handle). They use text-to-3D with the prompt "a red coffee mug with a handle." The result looks like a mug, but it's not their mug — the shape is different, the handle is wrong, the proportions are off. They don't realize that image-to-3D would have used a photo of their actual mug to generate an accurate 3D model.

Each 3D generation mode serves a different production need. Knowing when to use each demonstrates practical 3D generation engineering skills.

What you'll build
Implement text-to-3D, image-to-3D, and multi-view reconstruction using the Tripo API. Understand Gaussian Splatting for photorealistic 3D. Document when to use each mode.

3D Generation Modes

1. Text-to-3D

Generate a 3D model from a text description:

import requests

def text_to_3d(prompt: str, topology: str = "standard",
               format: str = "glb") -> dict:
    """Generate 3D model from text description."""
    response = requests.post(
        "https://api.tripo3d.ai/v1/generate",
        headers={"Authorization": f"Bearer {TRIPO_API_KEY}"},
        json={
            "type": "text_to_model",
            "prompt": prompt,
            "topology": topology,
            "format": format,
        }
    )
    return response.json()

Best for: concept exploration, rapid prototyping, creative assets Limitation: the model won't match a specific real-world object

2. Image-to-3D

Generate a 3D model from a single 2D image:

def image_to_3d(image_url: str, topology: str = "standard",
                format: str = "glb") -> dict:
    """Generate 3D model from a single image."""
    response = requests.post(
        "https://api.tripo3d.ai/v1/generate",
        headers={"Authorization": f"Bearer {TRIPO_API_KEY}"},
        json={
            "type": "image_to_model",
            "image": image_url,
            "topology": topology,
            "format": format,
        }
    )
    return response.json()

Best for: product visualization, e-commerce, accurate object reconstruction Advantage: the 3D model matches the actual object in the image

3. Multi-View Reconstruction

Generate a 3D model from multiple images taken from different angles:

def multi_view_to_3d(image_urls: list[str], format: str = "glb") -> dict:
    """Generate 3D model from multiple view images."""
    response = requests.post(
        "https://api.tripo3d.ai/v1/generate",
        headers={"Authorization": f"Bearer {TRIPO_API_KEY}"},
        json={
            "type": "multi_view_to_model",
            "images": image_urls,  # List of URLs from different angles
            "format": format,
        }
    )
    return response.json()

Best for: photorealistic 3D, accurate reconstruction, product photography Advantage: multiple angles provide depth and detail that single images can't

You need a 3D model of a specific product for an e-commerce site. Which mode should you use?

Both image-to-3D and multi-view are correct. Image-to-3D uses a single product photo (simpler, faster). Multi-view uses photos from multiple angles (more accurate, more detailed). For e-commerce, start with image-to-3D for speed. If higher accuracy is needed (e.g., customer can rotate the product in 3D), use multi-view with 3-5 angles. Text-to-3D won't match the specific product.

4. Gaussian Splatting

Gaussian Splatting represents 3D scenes as collections of 3D Gaussians (particles) rather than meshes:

# Gaussian Splatting is a different 3D representation
# It's not mesh-based — it's a radiance field

# Key difference:
# Mesh: vertices + faces (polygons)
# Gaussian Splatting: 3D Gaussians with position, rotation, scale, color, opacity

# Advantages:
# - Photorealistic rendering
# - Real-time rendering (100+ FPS)
# - Captures view-dependent effects (reflections, transparency)

# Disadvantages:
# - Not editable (can't modify geometry)
# - Large file sizes
# - Not compatible with game engines (need mesh for games)

Gaussian Splatting vs Mesh

Property	Gaussian Splatting	Mesh
Representation	3D Gaussians (particles)	Vertices + faces
Rendering	Photorealistic, real-time	Good, depends on textures
Editability	Not editable	Fully editable
File size	Large (100MB+)	Small (1-50MB)
Game engine	Not supported	Fully supported
Best for	Photorealistic visualization	Games, AR/VR, animation

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers Mode Comparison, Production Pipeline, Key Takeaways, What's Next — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.